Resillent By Design

How to build fault tolerant systems which can withstand failure and respond under load.

Transcript

1. RESILIENT BY DESIGN SMIT SHAH github.com/who828 twitter.com/who828
3. WHY DO WE CARE ABOUT RESILIENCE?
4. HOW DO WE BUILD A RESILIENT SYSTEM?
5. WE MUST DESIGN FOR RESILIENCE FROM THE START
6. RESILIENT DESIGN PATTERNS
7. RESILIENT DESIGN PATTERNS ▸ Fail fast ▸ Bounding ▸ Circuit breakers ▸ SteadyState ▸ Bulkheads
8. FAIL FAST
9. LITTLE’S LAW: L = Λ * W L = Length of the Queue λ = Message Arrival Rate W = Mean Time in System
14. BOUNDING
15. BOUNDING ▸ Use Timeouts
16. BOUNDING ▸ Use Timeouts ▸ Limit Memory/CPU
17. BOUNDING ▸ Use Timeouts ▸ Limit Memory/CPU ▸ Use Explicit Bounded Queues
18. CIRCUIT BREAKERS
20. BULKHEADS
23. STEADYSTATE
24. STEADYSTATE ▸ Avoid fiddling
25. STEADYSTATE ▸ Avoid fiddling ▸ Roll your logs
26. STEADYSTATE ▸ Avoid fiddling ▸ Roll your logs ▸ Have a data archiving strategy
27. Software design today only talks about what systems should do. It doesn’t address the things systems should not do. — Michael T. Nygard
28. PUTTING IT TOGETHER ▸ Fail fast ▸ Bound your resources ▸ Use circuit breakers ▸ Isolate failures
29. QUESTIONS?
30. REFERENCES Release It! The Art of Capacity Planning Thinking for Programmers Responding in Timely Manner Reactive Service Levels Queuing Theory Calculator